I don't quite get what they are currently doing. Is it that they have some kind of Internal mitm proxy that proxies every TLS session with external hosts, recording the conversation for later offline analysis? Why wouldn't they still be able to do that? Or are they just worried about their ability to intercept internal conversations? I'd be really interested to know what a typical bank security architecture looks like, does anyone have a reference for further reading perhaps?
Hmm, but why can't they just then record the session keys as before even for forward secure TLS (presuming we are talking about clients inside the bank initiating TLS sessions to external parties)?